From: Markus Mayer (mymaillists@gmx.at)
Date: Wed Nov 21 2007 - 07:47:09 EST
Hi all,
I'm having a hard time with Sun Cluster 3.2 in a two node cluster. We have a
Hitachi AMS500 and HDLM 5.9.0.0, Solaris 10 update 3, all patches until the
start of September installed. I have 30 disks from the storage array made
available, and to try to keep an overview of what disk came from which pool
on the array and for what purpose it was, I renamed the did devices on the
disks (eg: didadm -t d35:d101). The renaming procedure went through without
any problems, however since rebooting, I can do nothing with these did
devices any more. For example, the boot messages for one of these disks is:
Nov 21 13:12:48 wombat Cluster.CCR: /usr/cluster/bin/scgdevs: Could not
register disk
Nov 21 13:12:48 wombat Cluster.CCR: /usr/cluster/bin/scgdevs: Could not
stat /dev/global/dsk/d123s2:
Nov 21 13:12:48 wombat Cluster.CCR: No such file or directory
didadm -c or -R returns no errors, so it thinks everything is ok.
I have tried everything I can think of to access the disks, or to remove them
from the device tree, to rename them again to device numbers under 100, to
clean the did device tree, and so on, all with no success. (didadm -C
then -r, devfsadm -Cv, didadm -t d101:d35)
I have examined the problem and seen that there are no did device entries
(links) in /dev/did greater than d100, the devices go from did/d0 to
did/d100, there is nothing for did/d101 and onwards. The same applies
in /devices/pseudo/did. According to Sun documentation, these should however
be dynamically generated.
I cna see two possibilities here -
1. Find a way to get cluster to generate the missing device entries.
2. Find a way to remove all the problem entries and go back to something more
conservative.
So far I have not been able to find a way to do either of these. I have
considered the possibility of manually brutally deleting the troublesome
entries, however I don't know what consequences this will have.
I would be grateful to anyone who could point me in a direction to either get
cluster to delete the troublesome entries and start again, or generate the
missing entries.
Thanks
Markus
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:42:32 EDT